Today, many fossilized Megalodon teeth are still found and are often sold as collector\u2019s items. But what can X-ray tomography tell us about an ancient Megalodon tooth?<\/strong> Let\u2019s take a look.<\/p>\n\n\n\n The photo on the left was taken with a smartphone. The CT-scan was made with a high resolution transmission tube system<\/strong>, coupled with a 100\u00b5m pixel pitch detector<\/strong> for catching a good amount of details.<\/p>\n\n\n\n A couple of things are visible in the cross-sectional CT slices:<\/p>\n\n\n\n These 4 components can all be visualized as shown on the CT 3D View.<\/p>\n\n\n\n On the right is an example of a CT view of all segmented components in the Megalodon tooth. <\/p>\n\n\n\n The deep learning process<\/strong> is achieved using the Segmentation Wizard tool. Some input frames are provided and a 2.5D Unet model is trained based on this input. Some screenshots are shown in the figure on the right.<\/p>\n\n\n\n The typical work required to provide input in one or more input frames takes less than an hour, and the training process itself (with no operator dependance) can run for as long as 24 hours.<\/p>\n\n\n\n In the example (right image), the training was 10 hours in total. The speed of training depends on the computer specifications<\/strong> especially:<\/p>\n\n\n\n In this example, the dataset size was 8.5 Gb and the deep learning default parameters were used for a 2.5D Unet model on a machine with 64 Gb RAM and a Nvidia 3070 laptop GPU (8 Gb onboard memory).<\/p>\n\n\n\n Now, we can use the results to investigate each of the features in the tooth in more detail. <\/p>\n\n\n\n First up, let\u2019s take a look at the cracks<\/strong>. When viewing the cracks in 3D, there is a clear vertical orientation of the largest cracks from the base towards the tooth cutting edges. The visualization is color-coded<\/strong> by local thickness<\/strong> in the image on the right. The local thickness shows the cracks are mostly below 0.5 mm.<\/p>\n\n\n\n The root canal network visualized in 3D is shown in image on the left. By viewing this from the side it is best to see the channels connected to the holes to the front surface. The color coding indicates the local thickness<\/strong>, emphasizing the change in width<\/strong> of the channels. The connectivity in three dimensions is particularly interesting, as is the curvature of the large channels near the tooth base.<\/p>\n\n\n\n![\"\"](\"https:\/\/tetravision.be\/wp-content\/uploads\/2025\/05\/Megalodon-tooth-photo-1.avif\")
<\/a><\/figure>\n\n\n\n<\/a><\/figure>\n<\/figure>\n\n\n\nCapturing the details<\/h3>\n\n\n\n
\n
All segmented components<\/h3>\n\n\n\n
\n
<\/a><\/figure>\n\n\n\nDragonfly’s deep learning<\/h3>\n\n\n\n
\n
<\/a><\/figure>\n\n\n\nLocal thickness color mapping<\/h3>\n\n\n\n
<\/a><\/figure>\n\n\n\nRoot canal network in 3D<\/h3>\n\n\n\n
<\/a><\/figure>\n\n\n\nKey takeaways<\/h3>\n\n\n\n
\n